Skip to content

Latest commit

 

History

History
123 lines (72 loc) · 6.02 KB

03.md

File metadata and controls

123 lines (72 loc) · 6.02 KB

Overview
Overview		 Prerequisites
Prerequisites		 Design
Design		 Development
Development		 Optional
Optional

Lex Workhsop

Design

In this section you will create a Lex chatbot and then design the conversational user experience (UX). You will test the conversation UX before adding any business logic.

1. Create Lex Chatbot

Task: You will sign in to AWS console, access the Amazon Lex service, and create a new custom bot named "WhereBot". You will configure your bot to be a text-based chat bot. You will assign the bot to auto versioning.

Step-by-step instructions (expand for details)

  1. Sign in with your AWS developer account at https://console.aws.amazon.com. Under AWS services, type "Lex", then select Amazon Lex:

    aws console lex

  2. If this is your first time, select the Get Started button

  3. On the Create your bot screen, select Custom bot and apply the following values:

    • Bot name: WhereBot
    • Output voice: None. This is only a text based application
    • Session timeout: 1 min
    • COPPA: No

    aws lex create custom

    After selecting the Create button, you should now be in the WhereBot Lex service with Editor as the active tab.

  4. Select the Settings tab and confirm Aliases is the selected option. Apply the following values and then select the + button:

    • Alias name: Demo
    • Bot Version: Latest

    aws lex version

    Switch back to Editor tab. You have successfully created your Lex custom bot. Now it is time to design it.

2. Design the Conversational UX

Task: You will create a new Intent named GetLocationName and design it to answer a question such as: Where is Nic? You will test intent for conversational accuracy.

Step-by-step instructions (expand for details)

  1. In the Editor tab, select the + Create Intent button. In the pop-up dialogue box, select + Create intent option. Type "GetLocationName" and select Add button.

    aws lex create intent getlocationname

    After selecting the Add button, you should now see GetLocationName at the top of the main panel.

  2. Select the Slot types + option from left panel. In the pop-up dialogue box, select + Create slot type option. Type in the following for each field:

    • Slot type name: TrackablePerson
    • Description: Person tracked with geolocation
    • Slot Resolution: Restrict to Slot values and Synonyms
    • Value:
      • Nic: Nicolas
      • Mike: Michael
      • Jay: Jayson
      • Richard (no synonym)

    aws lex create slot trackableperson

    After selecting the Add slot to intent button, you should now see a new line in the Slots section with the name of slotOne and a type of TrackablePerson.

  3. Type "Who?" for the Prompt value for slotOne

    Because the Required setting is checked, a prompt is needed if the user does not initially provide it.

  4. Add the following values for Sample utterances (select + to add each value):

  • Where
  • Where is {slotOne}

aws lex save getlocationname

Sample utterances are how your bot will be invoked.

  1. Select the Save Intent button.

  2. Select the Build button at top of page and wait for success.

aws lex build

  1. Expand the Test bot panel on righ side of screen and type "Where" in the textbox with the watermark of Chat with your bot... When prompted for Who?, type "Nic".

aws lex test nic

The final response generated only returns the values of which Intent was invoked (GetLocationName) and what parameter (slotOne:Nic) was provided. This behavior is due to Fulfillment option for intent set to Return parameters to client. We will keep it at this setting for now to confirm testing of conversational UX. When the bot was invoked by typing "Where", it prompted for additional information because slotOne is required.

  1. Select Clear chat history and type "Where is Michael"

aws lex test nic

The bot was invoked by typing "Where is Michael" - which used the Where is {slotOne} utterance. Because the value for slotOne was provided, no additional prompts were needed. Also note that although "Michael" was typed in, the parameter type returned is "Mike", because of the synonym relationship.

  1. Select Clear chat history and type "Where is Fred". When prompted, type "Fred" again. When prompted again, type "Freddy"

aws lex test fred

The bot was invoked by typing "Where is Fred". However, because an unknown value was provided for slotOne, the bot repeatedly prompted Who? for a known value until the third failed attempt.

  1. Select Clear chat history and type "Howdy". When prompted, type "where is jayson" (all lower case)

aws lex test howdy

The bot was did not successfully invoke an intent because "Howdy" is not a recognized utterance. The bot kindly asked for another attempt. When entering the values in lower case, the bot still invokes the intent properly. Again, providing the value of "Jayson" returned "Jay" as slot value due to synonym relationship.

You have successfully tested the conversation UX.

You are now ready to proceed to the next step.